2013-10-16 10:19:39

by Hiroshi Doyu

Subject: [PATCH 1/1] ARM: dma-mapping: Drop GFP_COMP for DMA memory allocations

arm_iommu_alloc_attrs wants to split pages after allocation in order
to reduce the memory footprint. This does not work well with GFP_COMP
pages, so drop this flag before allocation.

ref: ea2e7057c0234cfb8b09467d8f137760d371fc72:

ARM: 7172/1: dma: Drop GFP_COMP for DMA memory allocations

dma_alloc_coherent wants to split pages after allocation in order to
reduce the memory footprint. This does not work well with GFP_COMP
pages, so drop this flag before allocation.

Signed-off-by: Hiroshi Doyu <[email protected]>
---
arch/arm/mm/dma-mapping.c | 7 +++++++
1 file changed, 7 insertions(+)

diff --git a/arch/arm/mm/dma-mapping.c b/arch/arm/mm/dma-mapping.c
index f5e1a84..955dd3e 100644
--- a/arch/arm/mm/dma-mapping.c
+++ b/arch/arm/mm/dma-mapping.c
@@ -1321,6 +1321,13 @@ static void *arm_iommu_alloc_attrs(struct device *dev, size_t size,
struct page **pages;
void *addr = NULL;

+ /* Following is a work-around (a.k.a. hack) to prevent pages
+ * with __GFP_COMP being passed to split_page() which cannot
+ * handle them. The real problem is that this flag probably
+ * should be 0 on ARM as it is not supported on this
+ * platform--see CONFIG_HUGETLB_PAGE. */
+ gfp &= ~(__GFP_COMP);
+
*handle = DMA_ERROR_CODE;
size = PAGE_ALIGN(size);

--
1.8.1.5

2013-10-16 10:41:08

by Will Deacon

Subject: Re: [PATCH 1/1] ARM: dma-mapping: Drop GFP_COMP for DMA memory allocations

On Wed, Oct 16, 2013 at 11:19:13AM +0100, Hiroshi Doyu wrote:
> arm_iommu_alloc_attrs wants to split pages after allocation in order
> to reduce the memory footprint. This does not work well with GFP_COMP
> pages, so drop this flag before allocation.
>
> ref: ea2e7057c0234cfb8b09467d8f137760d371fc72:
>
> ARM: 7172/1: dma: Drop GFP_COMP for DMA memory allocations
>
> dma_alloc_coherent wants to split pages after allocation in order to
> reduce the memory footprint. This does not work well with GFP_COMP
> pages, so drop this flag before allocation.
>
> Signed-off-by: Hiroshi Doyu <[email protected]>
> ---
> arch/arm/mm/dma-mapping.c | 7 +++++++
> 1 file changed, 7 insertions(+)
>
> diff --git a/arch/arm/mm/dma-mapping.c b/arch/arm/mm/dma-mapping.c
> index f5e1a84..955dd3e 100644
> --- a/arch/arm/mm/dma-mapping.c
> +++ b/arch/arm/mm/dma-mapping.c
> @@ -1321,6 +1321,13 @@ static void *arm_iommu_alloc_attrs(struct device *dev, size_t size,
> struct page **pages;
> void *addr = NULL;
>
> + /* Following is a work-around (a.k.a. hack) to prevent pages
> + * with __GFP_COMP being passed to split_page() which cannot
> + * handle them. The real problem is that this flag probably
> + * should be 0 on ARM as it is not supported on this
> + * platform--see CONFIG_HUGETLB_PAGE. */
> + gfp &= ~(__GFP_COMP);

Right, except that we *do* have hugepage support on ARM, so this probably
needs fixing properly.

Will